This paper reports lessons learned from reliability theory-based LRFD calibration for internal limit states for reinforced soil walls subjected to soil self-weight under operational conditions. The example of the ultimate pullout limit state for steel strip reinforced soil walls is used to demonstrate key issues. A unique feature of the general approach is the use of bias values that include the influence of model error and other sources of variability in input parameters on nominal load and resistance values. The paper shows how bias values can be used to: a) develop analytical load and resistance models that are statistically more accurate and avoid unwanted dependencies; b) select load factors satisfying a target exceedance value, and; c) compute resistance factors meeting a target probability of failure (reliability index value). The general approach has application to rigorous LRFD calibration of a wide range of geotechnical soil-structure design problems.